problems in the international financial system - · pdf fileproblems in the international...

OECD Journal: Financial Market Trends

Volume 2014/1

© OECD 2014

1

Problems in the international financial system

by

Adrian Blundell-Wignall and Caroline Roulet*

Since the 1980s OECD investment-saving correlations – as an inverse measure ofeconomic openness – indicate a very wide disparity of openness between the OECDand emerging market economies (EMEs) with an absence of open markets in thelatter. Given the increasing weight of EMEs in the world economy this pattern ofgrowth with disparity of openness is ultimately unsustainable. This approach todevelopment is not in the interests of EMEs in the post-crisis global environment.Various studies show how the absence of capital mobility inhibits developmentthough private sector capital expenditure at the firm level. This paper generalisesthose findings in a panel study, showing that in the period since 2008 the increasedpresence of capital controls is associated with highly significant negative effects onbusiness investment. It suggests that the world economy could be entering a moredangerous phase of potential instability that is not in the interests of either theadvanced or the emerging world. There is scope for better policies to encourage moreopenness; the OECD Codes of Liberalisation could be an effective tool for managingthe reform process.

JEL classification: C23, D22, E20, F21, F31, F38, F43, G01

Keywords: Capital controls, capital flows, exchange rate management, savings andinvestment, corporate capital expenditure, emerging market economies (EMEs)

* Adrian Blundell-Wignall is the Special Advisor to the OECD Secretary General on Financial Marketsand Acting Director of the Directorate of Financial and Enterprise Affairs (www.oecd.org/daf/abw).Caroline Roulet is an OECD economist and policy analyst. This work is published on theresponsibility of the Secretary-General of the OECD. The opinions expressed and argumentsemployed herein do not necessarily reflect the official views of OECD member countries. Thisdocument and any map included herein are without prejudice to the status of or sovereignty overany territory, to the delimitation of international frontiers and boundaries and to the name of anyterritory, city or area.

PROBLEMS IN THE INTERNATIONAL FINANCIAL SYSTEM

OECD JOURNAL: FINANCIAL MARKET TRENDS – VOLUME 2014/1 © OECD 20142

I. IntroductionThe opening up of OECD economies with respect to international trade and

investment in the post-War period changed the operation of the international financial

system helping to stimulate growth and to encourage international investment. Multi-

national enterprises emerged and moved across international boundaries more freely and

this brought with it not only new markets, but also an ability to locate investment,

production and employment in multiple locations where supply chains can be managed in

the most advantageous way for the profitability of the firm, and where the savings to fund

that investment could be allocated from sources quite separate from the domestic location

of the firm’s headquarters. While these developments brought many benefits, they also

brought with them a greater complexity of financial needs, going well beyond the evidently

increased demand for cross-border banking. These included, inter alia:

● New products that would facilitate hedging of exchange rate and credit default risks.

● Financial engineering to best match maturities required by savers and investors, and to

take advantage of different tax and regulatory regimes that bear on the costs of doing

business (for which swaps were very convenient).

● Mergers and acquisitions not only of businesses, but of stock exchanges and related

markets with global capabilities.

● New platforms and technological developments to handle the trading of new products

with volatile mark-to-market prices.

Policy makers in OECD countries accommodated these trends via financial

deregulation – eliminating international capital controls,1 moving towards auction

systems for selling sovereign debt and developing monetary policy that moved away from

sectoral quantitative and interest rate controls to operate instead via market-determined

financial prices. This freeing up of financial markets followed after the opening of goods

markets, particularly from the early 1980s, and in some respects was the necessary

counterpart of it.

The trends towards openness in OECD economies were not mirrored in emerging

markets. Capital controls have remained very strong, and these measures often support an

exchange rate regime that is heavily managed versus the US dollar. The opening up of the

OECD since the 1980s has certainly benefited economic development in emerging market

economies (EMEs). To the extent that exchange rates can be held below market-determined

levels via intervention, and inflation consequences can be mitigated by capital controls,

then an export-oriented trade and development model can be successful in the absence of

crises. As many of these countries are relatively poor compared to OECD countries, some of

these policies might be regarded as appropriate for the stage of development and a

sequencing of deregulation will follow (like the OECD countries in the post-War period). But

the overall global balance has to be right, since these policies do not come without costs:

both for the countries imposing them and for OECD countries where the distortions in

prices created may have played a role as one of the causes of the financial crisis.


OECD JOURNAL: FINANCIAL MARKET TRENDS – VOLUME 2014/1 © OECD 2014 3

This paper examines and provides evidence on a number of problems in the

international financial system which may have become more apparent in the post-

financial-crisis environment. Section II examines the evidence on the intervention and

capital control practices of EMEs and their impact on unstable carry trade loops and on

trends in savings-investment imbalances in the global economy. It raises the issue of

whether a “tipping point” may emerge in the direction of unsustainable growth patterns.

Section III examines the micro evidence of the impact of reduced capital mobility on

private investment in EMEs, first by summarising the case studies literature and then by

testing the findings there in a more generalised study using a panel of 4 780 firms over the

period 2004 to 2012. Finally, in section IV, the findings are summarised and a concluding

comment on policy directions is offered. All of the technical details and results may be

found in the Annex.

II. Some global impacts of the rise of emerging marketsThe US dollar has long served as the reserve currency in the global economy. When

countries intervene to fix or manage their currencies versus the dollar as a growth strategy,

they acquire US dollars and typically recycle these into holdings of US Treasury securities,

which are the most liquid security (with little political risk) in the global economy. The

systematic undervaluation of exchange rates in many of these countries has been a feature

of the world economy since the early 2000s, with most Asian countries running trade

surpluses with the USA which require ex-post net lending to the US to fund its deficits. The

system somehow has to absorb these large persistent capital inflows that not only affect

the configuration of exchange rates but also put downward pressure on US Treasury yields.

US monetary policy is forced to respond to the conditions created by these flows. For

most of the period since the tech bust high and seemingly intractable unemployment has

been combined with low inflation (itself in part due to the global supply shock coming from

the EME world and the fear of job loss in the West that kept wage demands low) to

encourage low policy rates and, since 2008, the use of unconventional monetary policy.

Figure 1 shows the holdings of US Treasuries by foreigners (mostly central banks), which have

accelerated since the crisis and now total something like USD 5.8 trillion. Also shown on the

chart is the US Federal Reserve’s own balance sheet, which has also risen sharply to around

USD 3.8 trillion, and its holdings of Treasuries which have risen to around USD 2 trillion.

These sharp increases in foreign and holdings of Treasuries work to hold interest rates

down, while the foreign exchange intervention behind the purchases is holding the dollar

up against EME currencies. The rise in the Federal Reserve’s own balance sheet puts

opposing downward pressure on the dollar, providing a focal point for the clash between

two fundamentally different approaches to policy in the world economy – an interaction

that is not stable:

● Low interest rates and quantitative easing (QE) weaken the dollar, compared to what it

would otherwise be. At the same time investors move out on the risk curve to generate

better returns. Carry trade outflows into EMEs generate more intervention which,

without capital controls, risks credit boom-bust cycles in those countries managing the

exchange rate. These pressures generate incentives for more capital controls. Allowing the

exchange rate to move up without intervention would avoid credit boom-bust cycles, but

EMEs are fearful of the effect of a rising currency on exporting firms. With currency

intervention this fear is replaced by the risk of the loss of monetary control in the “risk-on”



years, and the possibility of dollar funding drying up when money is pulled out of EMEs in

the “risk-off” phase of the global cycle.

● If the dollar rallies in the “risk-off” phase, where money is withdrawn from EMEs and

carry trades are reversed, there is no incentive for EMEs to reduce capital controls since

their currencies are under downward pressure – indeed controls on outflows of various

forms are likely to increase during a crisis.

Capital controls therefore become “embedded” and risk “ratcheting-up” over the risk

cycle, thereby further distorting relative prices in the global economy.

The impact of dollar-bloc investment in US Treasury securities

In the Annex A.I of the study a model of US 10-year Treasuries is estimated in a

co-integration and error-correction framework, based on: short rates; inflation; foreign

holdings of US Treasuries; and the Federal Reserve’s own holdings of bonds. In the long-run

model the main influences on the bond rate are inflation, short interest rates and holdings

of Treasury securities by foreigners and the US Federal Reserve. The contributions to the

overall fall in the US bond yield are shown in Figure 2:2

● The move up in foreign and Federal Reserve holdings of US Treasuries as a percentage of

GDP over the crisis period from December 2007 (21% of US GDP), to its peak (of about 46%)

by late 2013, amounts cumulatively to about 25% of US GDP. According to the model, with

other things given, this would be associated with a 3-percentage point fall in the US 10-year

bond rate (the largest proportion of which is due to foreign buying).

● The cut in the short rates over the same period was from 5.1% to 0.3%, or around

4.8 percentage points. Other things given, this would account for a 1.1 percentage point

fall in the equilibrium 10-year bond rate.

Figure 1. Foreign central bank holding of Treasuries & the Fed balance sheet

Source: US Treasury, Federal Reserve, OECD.

0

1 000

2 000

3 000

4 000

5 000

6 000

7 000

Dec. 9

8

Aug. 99

Apr. 00

Dec. 0

0

Aug. 01

Apr. 02

Dec. 0

2

Aug. 03

Apr. 04

Dec. 0

4

Aug. 05

Apr. 06

Dec. 0

6

Aug. 07

Apr. 08

Dec. 0

8

Aug. 09

Apr. 10

Dec. 1

0

Aug. 11

Apr. 12

Dec. 1

2

Apr. 14

Aug. 13

$bn

Other

Brazil

UK

Carribean

Korea

OPECJapan

Fed Balance Sheet

SEA

China

US FED Hld Tsy



● The fall in inflation, from around 4% to just less than 2% over this period, on average,

subtracted around 50 basis points from bond yields over the period.

Since US short rates have hit the zero-rate bound to their contribution by 2009 (see

Figure 2), most of the work in driving down bond yields has been carried by foreign and

(particularly recently) US Federal Reserve buying of US Treasury securities. In this context,

it is worth noting that the often-mentioned potential sell-off threat to US longer-term

yields that would be caused by any shift of foreign buyers to other currencies for reserve

management purposes need not have any negative impact on current monetary conditions

supporting the US economy – such a shift would help to weaken the currency, while the US

Federal Reserve could, if it so wished, substitute perfectly well for the foreign Treasury

buying component.

The low US bond yields analysed here reinforce the carry trade search for better yield,

including EME sovereign bonds and (more frequently) EME credit.

Carry trade feedback effects

These feedback effects are unstable for the global economy. Figure 3 shows the total

return index for the low-duration 1-3-year Treasury bonds, and the longer-duration

10-year-plus index. This is compared to the Merrill Lynch Emerging Market Credit Index

from late 2008, the worst point in the crisis. As the crisis hit in 2008-09, this initially led to

a safe-haven trade favouring US Treasuries, and a huge sell off in EME economy credit, as

capital flowed out of these countries in the “risk-off” trade. Subsequently, the policies

discussed above led to extremely low interest rates in the US and other OECD countries,

and inflows into emerging markets surged as the carry trade search for yield resumed with

added force. This resulted in one of the biggest rallies in EME credit in history: the move up

in the EME total return index shown in Figure 3 from January 2009 to February 2014 was a

124% gain, compared to a cumulative 30% return for the US 10y-plus index over that same

period, and an even much smaller return for lower-duration Treasuries. As noted above,

these flows generated strong foreign exchange market intervention over this post-crisis

Figure 2. Main contributions to US 10-year bond yield

Source: Authors’ calculations based on data from Thomson Reuters Datastream.

-4

-3

-2

-1

0

1

2

1998

M12

1999

M06

1999

M12

2000

M06

2000

M12

2001

M06

2001

M12

2002

M06

2002

M12

2003

M06

2003

M12

2004

M06

2004

M12

2005

M06

2005

M12

2006

M06

2006

M12

2007

M06

2007

M12

2008

M06

2008

M12

2009

M06

2009

M12

2010

M06

2010

M12

2011

M06

2011

M12

2012

M06

2012

M12

Contribution (%)to US 10Y Sov Bond Yield

Contribution of CPI_MOM Contribution of LIBOR_3M Contribution of TS_FHDL_GDP



period, the investment of which pushed US Treasury yields down further, reinforcing the

carry trade (and hence intervention and capital controls) over this period in an unstable

loop.

As monetary pressure built up during the “risk-on”phase, countries’ intervention

policies risked excess credit cycles and this resulted in a series of increased capital control

measures, some of which are set out in Table 1.

Figure 3. US Treasury versus emerging market credit USD: total return index


Table 1. Capital controls introduced

Country

Capital control measures introduced in EMEs since 2009

Restrictions on portfolio flows Restrictions on banking flows

Tax on foreign investments

Restrictions by asset type or maturity

Tax on short-term external borrowing

Quantitative limits on banks’ FX exposure

Required reserves on FX liabilities

Brazil Oct, Nov 2009 (R) Mar, Apr, Jul, Aug 2011

Jan, Jul 2011

Oct, Nov 2010 (R) Mar, Jun, Dec 2012 Dec 2012

Jul, Dec 2011 (R)

Indonesia Mar, Jun 2010 Jun, Dec 2010 Dec 2010 (R)

Apr 2011

Korea Nov 2010 (R) Jul 2011 Apr 2011 Nov 2009 (D), Jan 2010

Jan 2012 (R) Jun 2010 (D), Jun 2011 (D)

Nov 2012 (D)

Chinese Taipei Nov 2009 (R) Dec 2010 (D) Jan, Dec 2010 (R)

Nov 2010 (R)

Thailand Oct 2010 (R)

Note: “R” denotes measures that discriminate by residency and “D” denotes limits on banks FX positions.Source: Authors’ calculations from national press releases and media articles; Ahmed and Zlate (2013).

50

100

150

200

250

300

Nov. 0

8

Feb.

09

May. 0

9

Aug. 0

9

Nov. 0

9

Feb.

10

May. 1

0

Aug. 1

0

Nov. 1

0

Feb.

11

May. 1

1

Aug. 1

1

Nov. 1

1

Feb.

12

May. 1

2

Aug. 1

2

Nov. 1

2

Feb.

13

May. 1

3

Aug. 1

3

Nov. 1

3

Feb.

14

May. 1

4

Aug. 1

4

Nov. 1

4

IndexUS Tsy Tot Ret 1-3y US Tsy Tot Ret +10y Merill Tot Ret EM Credit



The joint effect of this is that the US economy cannot have the exchange rate regime

it needs against trade partners that are jointly larger than the US economy itself and are

growing rapidly. For the USA a stronger dollar versus the dollar-bloc (due to intervention)

and lower Treasury yields (dollar-bloc reserves investment) contributed to financial

imbalances that played a role in the build-up of imbalances before the crisis. The stronger

dollar works against manufacturing and other traded goods sectors and favours services

and housing, while the lower Treasury yields in relation to policy-determined short rates

feeds into the pricing of mortgages and other financial securities that spread into sub-

prime borrowers and securitisation processes. Nor do very low rates favour business

investment which, in larger firms, is based on retained earnings (and hence on the cost of

equity and accelerator mechanisms, discussed further below). More recently low interest

rates and the exchange rate policies of EMEs work in a feedback loop that helps boost the

carry trade and risks more financial market volatility later on.

Unbalanced economic development

Another analytical measure of financial and goods market openness is the correlation

between national saving and investment. Feldstein and Horioka (1980) interpret the high

correlation between these variables in the 1970s and 1980s to imply that global savings are

not sufficiently mobile to fund ex-ante demand for investment goods – or, alternatively, to

absorb excess national savings. Numerous subsequent articles have mostly corroborated

the findings and offered alternative explanations.3 For example, some argue that real

productivity and terms of trade shocks, or a fall in the rate of time preference that acts to

lift saving, may result in a high saving-investment (S-I) correlation in the presence of a

non-traded goods sector.4 If the non-traded goods sector is labour-intensive, a rise in

productivity would release factors of production, and proportionately more labour would

be allocated to the traded-goods sector. This would result in a higher marginal product of

capital in the traded-sector, thereby raising the desired capital stock, resulting in the co-

movement of savings and investment. These and other arguments may have some

application in developing countries, where covered interest parity does not hold, but this

should not be the case for most OECD countries.5

Within the borders of countries free trade does apply and capital can be assumed to be

perfectly mobile so that the S-I correlation should be zero. Helliwell (1998), using regional

data, shows there is little correlation between savings and investment in the Canadian

provinces, just as theory would predict, regardless of the presence of traded and non-

traded goods. This latter finding is quite interesting, and implies that the freeing up of

international trade and capital flows should also see something similar occurring in

countries and regional groupings. Countries which open up to foreign private participation

in domestic investment opportunities benefit from technology transfer, synergies in the

global supply chain, and resources development. Such countries should see S-I correlations

decline over time as this opening up occurs. Countries that are not open, or which are

excessively selective in their openness, should see higher more stable S-I correlations.

The S-I correlation in OECD countries has declined continuously in an extending

sample period from 1960 (from 0.7 1960-1964, to 0.3 from 1960 to 2011).6 To explore the case

for the EMEs the cases of China, India, Brazil, Mexico, South Africa, and Korea7 are

considered, and complete data is available for this group from 1982. Despite some modest

efforts at opening up at times, both the structure of trade and the maintenance of

significant capital controls appear to be consistent with a much higher S-I correlation than



found for the OECD. In an extending sample from 1982, EME correlations remain at around

0.7 and show no sign of declining over time. This suggests a very wide disparity of

openness between the OECD and EMEs in terms of willingness to open trade and to remove

capital controls.

Figure 4 shows the results for the 5-year rolling window of the S-I correlation. The

OECD countries reflect a sharp decline in the 2000s essentially to zero. The EMEs show no

evidence of decline. The coefficient for the 5-year period to 1986Q4 was around 0.78 and it

remains around that level in the most recent period.

The size of the EMEs in the global economy and the tipping point issue

Of course, what is true about individual countries and groupings of countries cannot

be true of the full global economy. Were it possible to have high-quality data for all

countries in the world, then savings and investment must be co-integrated. It is not

possible for the whole world to violate the budget constraint that savings equals

investment. If the non-OECD becomes large enough, a continuation of the recent S-I

correlation trends would in the end require OECD saving in aggregate to become negative

in order to support the growth strategy of the dollar bloc – a clearly unbalanced,

unsustainable and unlikely picture for the global economy.8

The intervention and capital controls process does not matter for the global economy

if a small country does this in an isolated way. In the early period of globalisation this

feature of the world financial system didn’t really matter. But since the end of the 1990s the

size of the BRICS and other countries that manage the exchange rate has risen strongly,

bringing with it a new set of issues. Figure 5 shows the PPP shares of the USA, the BRICS,

Japan, the EU, other Anglo-Saxon countries, and the rest (mainly other EME countries) in

world GDP. The US share is around 20%, whereas the approximate size of the BRICS and

other EMEs has risen from around 28% in 1980 to about 50% currently. This latter share is

expected to rise further.

Figure 4. Saving-investment correlation, EMEs vs. OECD, rolling window

Source: Authors’ calculations.

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1982

Q1-

86 Q4

1985

Q1-

89 Q4

1988

Q1-

92 Q4

1991

Q1-

95 Q4

1994

Q1-

98 Q4

1997

Q1-

01 Q4

2000

Q1-

04 Q4

2003

Q1-

07 Q4

2006

Q1-

10 Q4

All OECD Emergers



III. Are exchange rate management and capital controls still benefiting EMEs?Recent literature dealing with the effects of capital controls on economic well-being

show mixed results. One strand focuses on capital controls as a macro-prudential measure

as in Rey (2013), who attempts to show the existence of a global financial cycle driven by US

monetary policy. Floating no longer guarantees independent monetary policy, and the

author argues that “independent monetary policies are possible if and only if the capital account is

managed”. The demonstration of this cycle is the basis of some very unorthodox

conclusions which are wholly rejected by the current study. The author argues that the

right policies to deal with the “trilemma” (independent monetary policy, floating exchange

rates and capital controls) are probably to take actions directly aimed at the main source of

concerns, i.e. excessive leverage and credit growth. This is argued to require a convex

combination of macro-prudential policies based on aggressive stress-testing and tougher

leverage ratios. Depending on the source of financial instability and institutional settings,

the use of capital controls as a macro-prudential measure is required to enable

independent monetary policy. Unfortunately the set of correlations considered in this

paper does not justify such strong conclusions. Such policies may indeed be highly

damaging, with countries believing that macro-prudential measures may be used to avoid

or delay appropriate monetary policy actions and much needed structural reforms.9

Aoki et al. (2010) suggest that capital controls could have welfare-enhancing aspects if

credit markets are imperfect. They investigate how the adjustment to liberalisation of

international financial transactions depends upon the degree of domestic financial

development. Using a model with domestic and international borrowing constraints, they

show that when the domestic financial system is underdeveloped, capital account

liberalisation is not necessarily beneficial because total factor productivity (TFP) stagnates

in the long-run or employment decreases in the short-run. They suggest that government

policy, including allowing foreign direct investment, can mitigate the possible loss of

employment, but cannot eliminate it unless the domestic financial system is improved.

Figure 5. The rising share of BRICS and other non-OECD countries

Source: IMF, OECD.

0

10

20

30

40

50

60

70

80

90

100

1980 1990 2000 2008 2013 2018

Other BRICS Japan Other Anglo Saxon Economies EU USA



Similarly, Calvo et al. (2000) suggest that capital controls could be beneficial with

imperfect information in which financial globalisation may promote contagion effects.

Financial globalisation may strengthen contagion effects by weakening incentives for

gathering costly information and by strengthening incentives for imitating arbitrary

market portfolios. Numerical simulations show that these two frictions can be

quantitatively significant and may induce large capital outflows unrelated to country-

specific fundamentals in emerging markets. These findings raise the question of whether

globalisation is necessarily welfare improving, and suggest that capital controls may be

desirable because of the welfare costs induced by contagion.

Jeanne et al. (2010) analyse prudential controls on capital flows to emerging markets

from the perspective of a Pigouvian tax that addresses externalities associated with the

deleveraging cycle. It presents a model in which restricting capital inflows during boom

times reduces the potential outflows during busts. This mitigates the feedback effects of

deleveraging episodes, when tightening financial constraints on borrowers and collapsing

prices for collateral assets have mutually reinforcing effects. In their model, capital

controls reduce macroeconomic volatility and increase standard measures of consumer

welfare.

Finally, Bruno and Shin (2014) develop methods for assessing the sensitivity of capital

flows to global financial conditions, and apply the methods in assessing the impact of

macro-prudential policies introduced by Korea in 2010. Relative to a comparison group of

countries, they found that the sensitivity of capital flows into Korea to global conditions

decreased in the period following the introduction of macro-prudential policies and

controls on capital flows. However, these measures were introduced after the financial

crisis. Korea faced a major funding squeeze during 2008. During a “risk-on” period when

investment into EMEs is strong, it is quite likely that controls can help resolve “tri lemma”

issues; but during crisis periods they may be counterproductive (see below).10

Other macroeconomic studies have found at best ambiguous results on the effects of

capital controls on economic well-being. For example, Prasad et al. (2003) found no

significant relationship between openness and growth in per capita income between

countries, after controlling for initial endowments. Their survey of other studies shows

mixed results (though none found that liberalisation reduces growth). Similarly, Satyanath

and Berger (2007) , in a panel of 50 (mostly) EMEs, controlling for standard economic growth

determinants, found that there is no statistically significant linkage between capital

controls on inflows and lower average economic growth over the period 1995-2005, i.e. prior

to the global financial crisis (GFC).

In an influential IMF paper Ostry et al. (2010) using a probit model approach found that in

the post-global-crisis period the intensity of capital controls was negatively associated with

the fall in output; i.e. made things better. Blundell-Wignall and Roulet (2013) reproduced those

results and tested their robustness. They found that the results were not robust to a simple

stability test. If Latvia (with by far the biggest GDP drop in the crisis) is excluded from the

decile of countries with the largest fall in output and the next-worse country (Russia) is

included, the results no longer hold. The key coefficient is then insignificantly different

from zero. In short, these tests show that the IMF results cannot be used as a basis for

claiming some form of general support for the use of capital controls. In testing the same

IMF capital control measures with panel data on GDP growth, including data both before

and after the global crisis, while also controlling for macroeconomic factors (lagged



domestic GDP world GDP growth and lagged oil prices), they find quite unambiguous

results: that in the pre-crisis period capital controls support growth but that they have had

a very negative impact in the post-crisis period.

In contrast to the above macro studies, there is a considerable body of microeconomic

evidence showing that EMEs that lift capital controls do experience the positive benefits

predicted by economic theory. Particularly where firms do not have ready access to

international capital markets, lifting controls sees a reduced dependence on cash flows for

capital expenditure (Harrison, Love and McMillan, 2004). Similarly, property rights improve,

and newly “investible” firms see increased investment, rising stock prices and a fall in the

cost of equity (Chari and Henry, 2004). Forbes (2007) presents an interesting summary of

her own and other works looking at very specific case-study evidence on the effects of

capital controls:

● Their presence increases financing costs by reducing the amount of capital available to

domestic firms.

● The attempt to avoid capital control leads to inefficiencies and corruption aspects that

have real costs for the firm.

● Desai, Foley and Hines (2004) study the behaviour of US multinationals and find that

affiliates operating in low capital mobility countries are 10% more likely to remit

dividends to parent companies (rather that retain earnings to invest), and the distortion

to profitability caused by such controls is equivalent to a 24% increase in the corporate

tax rate. Capital controls increase borrowing costs and reduce the size of foreign

investment by multinationals by some 13-16% compared to what it would otherwise be.

● Case studies show that the initial “costs” of removing controls is usually found in the

protected rent-seeking domestic firms that have to adjust to global competition. But for

smaller firms, the evidence shows that stock prices rise (reducing the cost of equity), and

they find it easier to attract capital.

● In the case of Chile, Forbes (2003) found that the encaje caused investment to plummet in

smaller firms, and investment was also generally lower for larger firms.

Forbes wonders whether it is possible to generalise this case-by-case micro evidence.

Since most of this micro evidence focuses on investment, the rest of this section

attempts to see whether the propositions may be more general in a panel study of the

effects of capital controls on the capital expenditure decisions of 4 780 non-financial

companies in the MSCI index, covering most countries in the world economy (both with

and without capital controls). The technical aspects of this study are set out in full in

Annex A.II.

The capital controls measure

None of the above studies use a dynamic measure of capital controls, and none of

them distinguish between countries that have non-deliverable forward (NDF) markets,

which emerge when controls on capital are particularly strong. Countries that impose

capital controls experience deviations from covered interest parity (CIP), particularly at

times when capital inflow or outflow pressures are strongest. CIP compares the yields on

assets with the same duration issued in different countries and denominated in different

currencies but hedged to eliminate currency risk. In the absence of capital controls, the

yield on the foreign asset hedged in the forward market should be in line with the domestic

rate, except for (small) transactions costs. While it is true that in extreme crisis situations



market illiquidity can be a factor driving deviations, these are much smaller than for

capital-control countries and have only a temporary effect. The interest parity condition

from the viewpoint of a dollar investor is:

F/S-(1+R)/(1+R*)= ε (1)

Where: (F) is the forward rate, (S) the spot rate, R the dollar interest rate and R* is the foreign

rate and ε is a deviation. There is a substantial literature on deviations from CIP as they

pertain to capital controls.

Capital controls leading to deviations from CIP reflect pricing distortions as inflows

and outflows cannot clear in the normal manner. Sometimes, despite the constraints, the

market may still cope with trades, albeit inefficiently. However, in other cases capital

controls are so restrictive in the forward market that an NDF market develops to deal with

the desired trading of capital flows. NDF markets are located offshore and are structured

contractually to allow net settlement in US dollars. The development of NDF markets

implies the presence of tough capital controls. The tighter the restrictions on the domestic

forward market are, the larger the deviations between the two markets will be and the

more difficult it becomes to arbitrage between the NDF, spot and domestic forward

markets. The NDF market is essentially “mimicking” the deliverable market with basis risk;

i.e. using conversion exchange rates to compensate counterparties in US dollars for

hedging currency movements that can’t be delivered in local currency.

One problem in using CIP as a capital control measure is that it can move above or

below the transactions cost inertia band (including positive or negative deviations). For this

reason use of CIP as the basis for such measures requires a time-series modeling

transformation to reflect the degree of capital mobility. Generally speaking, movements to

greater openness should see covered differentials diminish and their variance decline over

time. In the literature a dynamic measure of capital mobility treats the residual ε in

equation (1) as a time-series model that captures the autoregressive structure in both the

mean and the variance: that is, the conditional variance estimated from a GARCH (1,1)

model gives an indication of dynamic capital mobility (see the Annex).11 In the face of

capital flow pressures countries with greater capital controls would exhibit greater

conditional variance of CIP than would more open economies. The time-series models of

conditional variance used in this study were estimated on daily data (where available) from

2004 to 2013, for both the deliverable forward (DF) market and the NDF markets, so that all

significant countries in the world economy are covered. Since the MSCI capital expenditure

data is available on a consistent basis annually, the conditional variance was re-estimated

for each calendar year and the annual averages were taken.

The resulting data are shown in Figure 6 for the case NDF markets in the country

shown. The DF markets for countries without capital controls have very small means and

conditional variances (see Annex Table A2) and therefore are not shown here. All of the

NDF countries saw a sharp rise in conditional variance during the crisis, which fell back to

(still large) pre-crisis levels subsequently. In the cases of Indonesia, India, Argentina and

Egypt, the conditional variance of CIP is typically very high and has been rising again,

suggesting less mobility in the post-2012 environment in these countries. The Latin

American group of Brazil, Chile, Peru, and Colombia have relatively more open economies,

according to this measure, while the other countries shown lie somewhere between the

two extremes.



Testing the effect of capital controls on corporate capital expenditure

A panel regression approach for firm-level capital expenditure is used for all the non-

financial companies in the MSCI index, using a co-integration and error-correction

framework. The results are based on an unbalanced panel of 4 780 publicly traded

companies from 55 countries and 9 sectors over the period 2004-12. All of the data and the

formal results are set out in the Annex. The model for capital expenditure by the non-

financial firm (k) depends on sales (s), the cost of equity (q), the corporate bond rate (r), the

output gap (y-y*) and the volatility of the firms stock price (s) to control for firm-specific

Figure 6. Conditional variances (shown as standard deviations) NDF CIPs


0

0.5

1

1.5

2

2.5

3

3.5

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Std Dev (%)

CHINA

KOREA

CHINESE TAIPEI

INDIA

INDONESIA

MALAYSIA

PHILIPPINES

0

1

2

3

4

5

6

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Std Dev (%)

ARGENTINA

CHILE

COLOMBIA

EGYPT

PERUBRAZIL

RUSSIA



risk. As noted above, capital expenditure by firms within the country jurisdiction should be

negatively impacted by the presence of binding capital controls: the cost of equity will be

higher (stock price lower), the cost of debt funding will be higher, multi-national affiliates

will reduce expenditure in such jurisdictions and they will remit dividends to parent

companies rather that retain earnings for capital expenditure to a greater degree than

would otherwise be the case. The conditional variance of the CIP measure of capital

mobility (cv) is then added to this model in three alternative ways:

1. For more open economies that have no NDF market the conditional variance for the DF

market is used. For countries where an NDF market exists (regardless of whether there

is also a limited DF market) the conditional variance of this market is used. The sample

therefore includes countries with both heavier and lighter capital controls.

2. For all countries with a DF market, so that the conditional variance of CIP is included

only for countries with moderate or no capital controls.

3. For only those countries with an NDF market. This sample includes only countries with

relatively heavy capital controls.

The model tested is:

(2)

Where i = 1, 2, or 3, for the three above alternative specifications.

The results are set out in Table A4.1 of the Annex of this paper. All of the standard

macro arguments of the capital expenditure model (cost of equity and debt, the output gap

and risk) have the correct signs and are mostly supported by the data.12 For the dynamic

capital control variable experiments the following results are obtained:

1. For the full model with DF only countries and NDF countries (where one exists), the

coefficient is negative and significant at the 1% level. A 1-standard deviation rise in the

conditional variance would on average be associated with in a fall of company capital

expenditure equivalent to some 0.9% of total sales.

2. If DF market only countries are included, the conditional variance measure of capital

mobility is not significant.

3. If only countries with an NDF market are included (a much smaller sample), the

coefficient on the conditional variance term is negative and significant at the 5% level. A

1-standard deviation rise in the conditional variance would be associated with a fall in

company capital expenditure equivalent to 1.2% of total sales.

These results are consistent with earlier case studies that capital controls have

general negative effects on firm capital expenditure. Capital controls pose a real cost to

companies within such jurisdictions, resulting in less capital expenditure than would be

the case with more open markets with greater opportunities for lower cost funding and

better incentives to retain earnings for investment.

Testing the effect of capital controls before and after the crisis

Table A5 shows the result of a Wald Test for equation 2, for the case of capital mobility

measured by the conditional variance of the DF market for countries without an NDF

market and by that for the NDF if one exists (i.e. the i = 1 case above). The dummy variable

is set equal to 1.0 for the period 2008 to 2012, and zero elsewhere.

ks

a b q b r b y y b s b cvi= + ( ) + ( ) + −( ) + ( ) + ( )1 2 3 4 5*



(3)

The results show that b5 is positive but insignificantly different from zero, while the

coefficient b6 is equal to -1.29 and is significant at the 1% level. That is, post the crisis, a

1-standard deviation rise in capital immobility will be associated on average with capital

expenditure falling by 1.29% of company sales. The test of the null hypothesis that

b5 + b6 = 0 is rejected at the 1% level and the sum of the two coefficients is similar to that

obtained in estimating equation (2) without the dummy variable.

These before- and after-the-crisis results for firm-level capital expenditure are broadly

consistent with those found in the earlier study of the macroeconomic GDP effects of

capital controls using the IMF measure.13

● In the years prior to the recent crisis, capital controls appear to have had no significant

negative effects on investment. This is likely because company profits are supported by

the lower-than-otherwise exchange rate, and companies are not constrained for finance

with good earnings in the stronger growth years.

● However, since the crisis the exact opposite is found. Capital controls tend to become

stricter in EMEs, capital is less mobile and this constrains firm-level capital expenditure

even after controlling for macro factors and firm specific risk. Just at the time when

foreign capital is most needed, countries with the most controls appear to suffer the

greatest retreat of foreign funding and negative impacts on firm level capital

expenditure.

The overall effect using both pre- and post-crisis data (as noted earlier), is negative.

IV. Summary and conclusions on the requirements for sustainable global growth

This paper has considered evidence that raises serious concerns about the functioning

of global capital markets. At the international finance level many EMEs (’the dollar bloc’)

manage their currencies against the US dollar reserve currency. This requires intervention

in the face of capital inflow during “risk-on”periods, which would otherwise cause EME

currencies to appreciate. To prevent excess credit cycles and increased dependence on

dollar funding resulting from such exchange rate policies, these countries frequently

utilise capital control measures. Where these are very strict, NDF markets develop to

mimic the DF market by compensating hedgers and speculators in US dollars. The

proceeds of the exchange market intervention are typically invested in US Treasuries, and

this has been a major contributor to the huge build-up in these holdings to some 46% of US

GDP (including the Federal Reserve holdings). This build-up has made a very large

contribution to driving down the yields on longer-duration Treasury securities. The huge

rally in the Treasury market drove yields to such a historically low level that the carry trade

into EME credit since 2009 (in search of much higher risk and better yields) was reinforced.

The increased capital inflow pressure in EMEs led to more capital controls being

introduced.

World saving-investment correlations for the OECD and EMEs were also examined.

This correlation has fallen to zero for OECD countries in the most recent periods, and is in

strong contrast to the high correlation of above 0.7 for a selection of large EMEs. This is

ultimately unsustainable, given the global budget constraint that savings equals

investment. If investment-driven growth funded by high domestic saving remains the

ks

a b q b r b y y b s b cv b DUM cvi i= + ( ) + ( ) + −( ) + ( ) + ( ) + ( )1 2 3 4 5 6*



preferred growth strategy of EMEs, then as the EME world grows to be even larger than the

OECD a “tipping-point” problem begins to emerge. OECD savings cannot become

increasingly negative to support such a strategy in EMEs in the longer run.

A survey of micro case studies shows a multiple set of reasons why capital controls are

not in the interest of EMEs hoping to develop with private sector corporate investment as

opposed to state-driven SOE activity. A study of the impact of capital mobility (measured

by the conditional variance of CIP) on corporate capital expenditure of the non-financial

MSCI companies is shown to provide more generalised support for this case-study

evidence. Reduced capital mobility post the crisis (after controlling for the usual macro

determinants of investment) has had a negative impact on firm level capital spending.

While some caution is required, the findings of this study suggest that the world

economy could be entering gradually into a more dangerous phase of potential instability

that is not in the interests of either the advanced OECD or the emerging world. There is

scope for better policies to sequence a reform of the international financial system, moving

it away from the dangers of an increasingly large dollar-bloc focused on managed exchange

rates supported by capital controls. This scope arises because there is no conflict in policy

objective here. Reform would help to: reduce feedback loops that re-enforce the carry trade;

allow EME consumers to benefit from cheaper imports from abroad; permit more

domestic-demand-driven growth in EMEs to restore more balanced global growth; and

facilitate more efficient and cheaper funding for private capital expenditure in EMEs.

The OECD Codes of Liberalisation provide a flexible framework for dealing with a

gradual global liberalisation process, by allowing for specific “reservations”on becoming an

adherent (in the recognition of the fact that the country may not be in the position to

remove controls on specific operations), and incorporating principles of “standstill” (no new

controls, though derogations for temporary crisis measures can be accommodated). The

Codes are particularly useful for dealing with some of the key global issues by calling for:

the avoidance of discrimination between residents and non-residents; freedom of

residents to transact abroad, where national rules do not apply; and freedom of non-

residents to carry out certain operations in the territory of a Member, indistinctively of

treatment granted to residents (e.g. operations in foreign exchange).

Use of the Codes as a reform instrument would be preferable to risking more

disruptive outcomes in the future. The Codes have mechanisms providing for

transparency, accountability and peer review of remaining controls. Their purpose is to

minimise potential harm to other countries and to the collective interest in an open

financial system, and to help countries find less restrictive alternatives to the control

mechanism in question.

Notes

1. For the purposes of this paper, “capital controls” are defined as measures affecting operations inforeign currencies or with non-residents that have the effect of directly restricting capital flows.

2. The significant error-correction result establishes the presence of a co-integrating relationshipbetween the bond rate and the explanatory variables, and the lag structure demonstrates causality.It suggests a 7-month lag for the full effects to work through from a change in the co-integratingvariables to the observed 10-year bond rate.

3. The correlation itself is not due to econometric anomalies, such as the treatment of theendogeneity via instrumental variables. Feldstein and Bacchetta (1991) respond to criticisms of aneconometric nature, such as omitted variables (e.g. economic growth; Obstfeld, 1986), and dynamic



effects such as Summers and Carroll (1987) that governments adjust their budgets to “fill in” forprivate investment savings gaps rather that see capital inflows or outflows.

4. Murphy (1986), Engel and Kletzer (1989), and Wong (1990).

5. Obstfeld and Rogoff (2001) show that in a world with transport and transactions costs, combinedwith inter-temporal consumption smoothing, the S-I correlation can be positive.

6. See, Blundell-Wignall, Atkinson and Roulet (2013).

7. Korea and Mexico are included here as they joined the OECD only in the 1990s, and the focus of thispaper is historical.

8. If EME investment and saving correlation is represented by: Ie = a + 0.7Se + εe and OECD by:Io = b + εo, as implicit in the correlations shown here, then So = a + b + εe + εo – 0.3Se. A significantlylarge EME world would become destabilising if the correlations did not begin to change.

9. For example, Professor Rey focuses on the ‘leverage’ of banks which she finds is correlated with theVIX index, which is also correlated with the Fed funds rate (with lagged responses). Leverage istraditionally defined as assets divided by bank capital, but this is not Professors Rey’s definition.She divides assets by bank deposits. But if assets is denoted A, wholesale funding WF, deposits D,and equity E, the basic identity is A=D+WF+E. Dividing by D gives A/D=1+(WF+E)/D. The right-handside is akin to the ratio of unstable to stable funding. This “funding ratio” analogue is then takenfor 20 European countries and the median of these 20 is taken as ’the key variable’ for a study overthe period 1990 to 2012. This period corresponds with the sharp rise in the use of derivatives andalso contains the 3 biggest movements in the VIX index: the 2001 tech bust, the global crisis in2008-09 and the euro crisis in late 2011. Professor Rey finds that a rise in the VIX leads to a jumpup in A/D, followed by a sustained fall over a number of quarters. But this is what one would expectmechanically without a VAR study, since Professor Rey has constructed her data that way. A rise inthe VIX instantly raises the gross market value of derivatives, which is in the numerator butdoesn’t affect the deposits in the denominator. Following this initial impact during high-riskperiod, investors are induced to switch out of risk assets into cash deposits and at the same timelending in the interbank wholesale market stops. These are standard risk prevention reactions.The A/D ratio will fall over a prolonged period following a crisis. This comes from data definitions,and in no sense can her correlations justify her strong policy conclusions with respect to capitalcontrols. There are major problems with banks’ use of derivatives and their funding methods, butthese need to be addressed by very different policies than the imposition of capital controls.

10. It is worth noting that Australia, which had higher external funding of its banks than Korea, andhas no capital controls, experienced no crisis or problems during 2008-09.

11. See Faruqee (1992), Huang and Guo (2006), and Hutchinson, Kendall, Pasricha and Singh (2009).

12. The error-correction terms (see Annex) imply co-integration and strong causality, with a relativelyrapid mean lag of around half of one year.

13. See Blundell-Wignall and Roulet (2013).

References

Ahmed, S. and A. Zlate (2013), “Capital Flows to Emerging Market Economies: A Brave New World”,International Finance Discussion Papers, No. 1081, Board of Governors of the Federal Reserve System.

Blundell-Wignall A. and C. Roulet (2014), “Capital controls on inflows, the global financial crisis andeconomic growth: Evidence for emerging economies”, OECD Journal: Financial Market Trends,Vol. 2013/2, http://dx.doi.org/10.1787/fmt-2013-5jzb2rhkgthc.

Blundell-Wignall, A., P.E Atkinson and C. Roulet (2013a), “Integration versus Interdependence andComplexity in Global Trade and Finance in the Post-War Period”, in 50 Years of Money and Finance:Lessons and Challenges, Balling, M. and E. Gnan (eds), SUERF 50th Anniversary Volume, Larcier,Vienna.

Bollerslev, T. (1986). “Generalized Autoregressive Conditional Heteroscedasticity”, Journal ofEconometrics, Vol. 31 No. 3 (April), pp. 307-328.

Chari, A. and P.B. Henry (2004), “Is the Invisible Hand Discerning or Indiscriminate? Investment andStock Prices in the Aftermath of Capital Account Liberalisations”, NBER Working Paper, No. 10318.

Desai, D., F. Foley and J. Hines (2004), “Capital Controls, Liberalizations and Foreign Direct Investment”,NBER Working Paper, No. 10337.



Engle and Granger (1987), “Cointegration and Error Correction: Representation, Estimation andTesting”, Econometrica, 55, pp. 251-276.

Engel, C. and K. Kletzer (1989), “Saving and Investment in an Open Economy with Non-traded Goods”,International Economic Review, Vol. 30(4), pp. 735-52, November.

Faruqee, H. (1992), “Dynamic Capital Mobility in Pacific Basin Developing Countries: Estimation andPolicy Implications”, IMF Staff Papers, Vol. 39, No. 3 (September), Washington, DC.: InternationalMonetary Fund.

Feldstein, Bacchetta, (1991), “National Saving and International Investment”, in National Saving andEconomic Performance, Bernheim and Shoven (eds), University of Chicago Press.

Feldstein, Horioka, (1980), “Domestic Savings and International Capital Flows”, The Economic Journal,Vol. 90 (June), pp. 314-329.

Forbes, K.J. (2007), ”One Cost of the Chilean Capital Controls: Increased Financial Constraints forSmaller Traded Firms”, NBER Working Paper, No. 9777.

Forbes, K.J. (2007), “The Microeconomic Evidence on Capital Controls: No Free Lunch”, in SebastianEdwards (ed), Capital Controls in Emerging Economies: Policies, Practices and Consequences, University ofChicago Press.

Granger (1983), Co-integrated Variables and Error-Correcting Models, unpublished discussion paper, pp. 83-13,University of California, San Diego.

Hassett, K.A. and R.G. Hubbard (2002), “Tax policy and business investment”, Chapter 20 in Auerbach, A.J.and M. Feldstein (eds.), Handbook of Public Economics, Vol. 3, Elsevier.

Helliwell, J.F. and Ross McKitrick (1999), “Comparing Capital Mobility Across Provincial and NationalBorders”, Canadian Journal of Economics, Canadian Economics Association, Vol. 32(5), pp. 1164-1173(November).

Huang, Y. and F. Guo (2006), “Covered Interest Parity and Market Volatility: Asian Evidence”, Advancesin Financial Planning and Forecasting, Vol. 2.

Hutchison, M.J. Kendall, G.K. Pasricha and N. Singh (2009), “Indian Capital Control Liberalisation:Evidence from NDF Markets”, MPRA Paper 13630, University Library of Munich, Germany.

Murphy, R.G. (1986), “Productivity Shocks, Non-traded Goods and Optimal Capital Accumulation”,European Economic Review, Vol. 30(5), pp. 1081-1095 (October).

Obstfeld, M. (1986), “Capital Mobility in the World Economy: Theory and Measurement”, Carnegie-Rochester Conference Series on Public Policy, Elsevier, Vol. 24(1) (January), pp. 55-103.

Obstfeld, M. and K. Rogoff (2001), “The Six Major Puzzles in International Macroeconomics: Is There aCommon Cause?” in Bernanke, B.S. and K. Rogoff (eds.), NBER Macroeconomics Annual 2000, Vol. 15,pp. 339-412, National Bureau of Economic Research, Inc.

Ostry, J.D., A.R. Ghosh, K. Habermeier, M. Chamon, M.S. Qureshi and D.B.S. Reinhardt (2010), “Capitalinflows: the role of controls”, IMF Staff Position Note 10/04, Washington, DC.: International MonetaryFund.

Prasad, E., K. Rogoff, S.-J. Wei and M.A. Kose (2003), “Effects of Financial Globalisation on DevelopingCountries: Some Empirical Evidence”, IMF Occasional Paper, No. 220, Washington, DC.: InternationalMonetary Fund.

Satyanath, S. and D. Berger (2007), “Capital controls, political institutions, and economic growth:A panel and cross country analysis”, Quarterly Journal of Political Science, 3, pp. 307-324.

Summers, L. and C. Carroll (1987), “Why Is U.S. National Saving So Low?”, Brookings Papers on EconomicActivity, Vol. 18(2), pp. 607-642.

Wong, D.Y. (1990), “What Do Saving-Investment Relationships Tell Us About Capital Mobility?”, Journalof International Money and Finance, Vol. 9, pp 60-74.



ANNEX

Empirical results

A.1. US bond yield model and foreign holdings effectFigure A1 shows the Engle-Granger co-integration and error-correction model

estimates for the 10-year US Treasury bond yield,1 where CPI is the consumer price index,

LIBOR is the US 3-month rate, and TS_HLD_GDP is the holdings of Treasury securities by

foreigners and by the US Federal Reserve (shown in Figure 1 of the text).

The CPI, the short rate and the foreign holdings of US Treasuries are major factors in

the co-integrating equation. The Federal Reserve holdings have the right sign, but the size

of the effect (if present at all) is very small.

Table A1. US 10-year bond yield model

[1] [2]

CPI_MOM 0.22 *(1.88)

-

LIBOR_3M 0.22 ***(9.98)

-

TS_HLD_GDP -0.09 ***(-15.03)

-

RESID_EQ1(-1) - -0.13 ***(-3.57)

D_CPI_MOM - -0.03(-0.59)

D_LIBOR_3M - 0.15 **(2.26)

D_TS_HLD_GDP - -0.01(-0.23)

C 0.05 ***(32.17)

-0.0001(-0.64)

R2 0.83 0.08

F-Stat 263 4

Prob(F) 0.00 0.01

Durbin Watson Stat. 0.28 1.58

Total Obs. 169 168




A.II. Capital expenditure of non-financial companies in the MSCITo investigate the determinants of firms’ (excluding financials) capital expenditure (k),

a fairly standard model is postulated, incorporating microeconomic and macroeconomic

variables. The dependent variable is company’s capital expenditure which is measured as

a share of the firms’ total sales ( ).

The explanatory variables are:

● Cost of equity (q) which is measured as the dividend yield of the company plus the trend

rate of growth of earnings over the period 2000-12. A higher cost of equity discourages

firms to retain earnings for the purpose of capital expenditures. ( ) is expected to have

a negative relationship with this variable.

● Cost of debt (r) which is the yield of AAA-rating corporate bond index of the country

location of the company. A higher bond yield discourages firms from borrowing to

increase their capital expenditures. ( ) is expected to have a negative relationship with

this variable.

● Output gap (y-y*) which is measured as actual GDP minus potential GDP (estimated

using the Hodrick-Prescott Filter) as a percentage of potential GDP. A boom in economic

activity increases the propensity of firms to invest. ( ) is expected to have a positive

relationship with this variable.

● The 1-year volatility of national equity benchmark (s) is calculated from the standard

deviation of day-to-day logarithmic historical price changes. The 260-day price volatility

equals the annualised standard deviation of the closing price change for the 260 most

recent trading days. Higher market volatility implies a higher firm-specific risk premium

and discourages firms to retain earnings for the purpose of capital expenditures. ( ) is

expected to have a negative relationship with this variable.

● The 1-year average conditional volatility of Covered Interest Parity (CIP) which is the

conditional standard deviation calculated using a GARCH (1,1) model.2 Along with

interest parities, the conditional variance of CIPs might be a measure of dynamic capital

mobility. Indeed with greater capital mobility, not only covered differential rates but also

the variance would decline over time. CIP is calculated using deliverable and non-

deliverable forward rates considering 3 different possibilities:

1. CIP is calculated using deliverable forward (DF) rate for countries with only a DF

market. CIP is also calculated using non-deliverable forward (NDF) rate for countries

with both DF and NDF markets and for countries with only a NDF market (CV).

2. CIP is calculated using the DF rate. It means that countries with only a DF market and

countries with both a DF and an NDF market are included (CV_DF).

3. CIP is calculated using the NDF rate. It means that countries with only an NDF market

and countries with both DF and NDF markets are included (CV_NDF).

● The greater the volatility, the more CIP is deviating from the 0 equilibrium. This

phenomenon is observed in countries with strong capital control measures. Stronger

controls on capital flows discourage firms to retain earnings for the purpose of capital

expenditures. ( ) is expected to have a negative relationship with this variable.

● To investigate the possible changes of the impact of CIP deviations following the 2008

financial crisis, a dummy variable is interacted with the 1-Year average conditional

k

s

k

s

k

s

k

s

k

s

k

s



volatility of CIP in the CV regressions. This variable (DC) is equal to 1 after the most

recent financial crisis, i.e. from 2008 to 2012, and is otherwise 0.

Estimation method and data

When economic theory suggests – as in the case of investment models – that there

exists an equilibrium relationship between integrated variables, co-integration and vector-

error-correction models provide an efficient estimator to describe short-run dynamics.

Beyond these pure econometric reasons, co-integration techniques and VECM are

especially useful in modelling investment. As Hubbard and Hassett (2002) argue, a number

of fundamental variables affecting investment move together over the business cycle. This

causes simultaneity problems. Shifts in the investment function (often associated with the

business cycle) imply a positive relationship between the user cost of capital and

investment, while interest rate shocks cause negative correlation between the two. If the

first dominates, the user cost elasticity will be small, and accelerator effects (the impact of

output on investment) large. This problem could be circumvented by focusing on the long-

run relationship instead. In addition, VECM allows a rich dynamic representation of the

data, which often turns out to be very useful given the sluggish nature of capital stock

adjustment and investment. In this paper Engel and Granger methodology is used.

The econometric analysis is run on a global sample of 4 780 publicly traded companies

from 55 advanced and emerging countries and 9 sectors3 over the period 2004-12. Annual

market and macroeconomic data were extracted from Datastream. Company data were

also extracted from Bloomberg. Table A2 presents some general descriptive statistics of the

final sample. Table A3 contains correlation coefficients among all explanatory variables.

Results

Regressions are run on the global panel of non-financial companies for the years 2004

to 2012, involving some 47 149 observations. The results are shown in Table A4.1 and

Table A5.

Table A2. Summary descriptive statistics of the sample

Mean Median Maximum Minimum Std. Dev. Observations

k/s 9.5 4.7 99.8 -24.7 13.7 61365

q 17.8 16.5 203.6 -99.2 30.6 57324

r 4.8 5.0 15.8 0.5 2.6 60973

s 21.7 20.2 72.9 6.9 9.0 62140

y-y* 0.0 -0.2 17.8 -17.2 2.4 62140

cv 0.2 0.0 6.1 0.0 0.4 50933

cv_df 0.1 0.0 6.1 0.0 0.3 50762

cv_ndf 1.1 1.0 5.4 0.1 0.8 4261

Note: OECD (2004-12). All variables are expressed in percentage.Source: Authors’ calculations based on data from Bloomberg and Thomson Reuters Datastream.



Table A3. Correlations among the main explanatory variables

q r s y-y* cv

q 1

r 0.04 1

0.00

s 0.01 0.10 1

0.02 0.00

y-y* 0.00 0.05 -0.02 1

0.49 0.00 0.00

cv 0.03 0.40 0.04 0.01 1

0.00 0.00 0.00 0.07

Note: All variables are expressed in percentage. Correlation coefficients are calculated using a balanced sample.Figures in italics indicate values of the T-statistics that test for null hypothesis of Pearson’s coefficients of correlationequal to 0.

Table A4. The determinants of investment of non-financial companies

Cointegration equation Error correction equation

Dependent variable: CAPEX %SALES Dependent variable: Δ CAPEX %SALES

[1] [2] [3] [1] [2] [3]

q -0.19 ***(-8.97)

-0.20 ***(-9.23)

-0.23 ***(-3.04)

Resid_VECM(-1) -0.54 ***(-117.62)

-0.53 ***(-117.05)

-0.58 ***(-20.41)

r -0.12 ***(-3.28)

-0.07 *(-1.78)

-0.29 ***(-2.86)

Δ q -0.06 ***(-3.60)

-0.06 ***(-3.22)

-0.14 *(-1.78)

y-y* 0.08 ***(3.17)

0.07 ***(2.65)

0.06(0.71)

Δ r -0.07 *(-1.76)

-0.05(-1.29)

-0.01(-0.05)

s -0.02 **(-2.04)

-0.01(-1.48)

-0.07 ***(-3.24)

Δ y-y* 0.17 ***(6.75)

0.14 ***(5.65)

0.97 ***(6.46)

cv -0.88 ***(-2.55)

- - Δ s -0.02 ***(-2.44)

-0.02 ***(-2.47)

0.03(1.40)

cv_df - 0.31(0.55)

- Δ cv -0.43(-1.36)

- -

cv_ndf - - -1.18 **(-2.19)

Δ cv_df - 0.52(1.20)

-

C 0.13 ***(29.25)

0.13 ***(27.62)

0.19 ***(10.86)

Δ cv_ndf - - 0.07(0.11)

C -0.001 ***(-3.85)

-0.001 ***(-3.66)

-0.001(-0.43)

R2 0.74 0.74 0.70 R2 0.32 0.31 0.41

F-Stat 26.84 28.33 8.85 F-Stat 3.91 4.01 1.31

Prob(F) 0.00 0.00 0.00 Prob(F) 0.00 0.00 0.00

Durbin-Watson Stat. 1.04 1.02 1.25 Durbin-Watson Stat. 1.96 1.95 2.58

Total Obs. 45 576 45 689 3 193 Total Obs. 40 191 40 758 1 884

Note: This table shows the results of estimating a VECM model for an unbalanced panel of 4 780 publicly tradedcompanies from 55 countries and 9 sectors over the period 2004-12. Cross-section and time fixed effects are used inthe regressions as is the White cross-section covariance method. *, ** and *** indicate statistical significance at the10%, 5% and 1% levels, respectively.



Notes

1. The Phillips-Perron (PP) unit root test is employed, which corrects, in a non-parametric way, anypossible presence of autocorrelation in the standard ADF test. Tests are showing that all variablesare non-stationary series at levels but stationary series at first differences.

2. If we denote y generically as the CIP, a time-series model that captures the autoregressive (AR)structure in both the mean and the variance can be written as:

Where is the ARCH term (the squared error term in the previous time period) of q order,generally being news about volatility from the previous period; is the GARCH term (theconditional variance in the previous time period) of p order. Thus, yt follows an AR (h) process witha conditional variance equation described by a GARCH (p, q) process. The GARCH model isimplemented via maximum likelihood estimation of the log-likelihood function. The estimatedconditional variance will give us an indication of the evolution of capital mobility. In this paper,GARCH (1, 1) model is adopted, which is sufficient to capture the dynamics of the conditionalvariance of y. The properties of the dataset are examined before the analysis of the empiricalresults. The Phillips-Perron (PP) test is employed. It is found that the null hypothesis of one unitroot can be rejected in all of the time series. CIP time series are driven by AR (1) processes. It is alsoidentified that the estimated coefficients are significant and all the diagnostic statistics arereasonable.

3. These sectors are: oil and gas, basic materials, industrials, consumer goods, healthcare, consumerservices, telecommunication, utilities, technology.

Table A5. The determinants of investment of non-financial companies testing the impact of the 2008 financial crisis

Cointegration equation

Dependent variable: CAPEX %SALES

[1]

q c(1) -0.19 ***(-9.00)

r c(2) -0.10 ***(-2.54)

y-y* c(3) 0.08 ***(3.24)

s c(4) -0.02 ***(-2.42)

cv c(5) 0.47(0.83)

cv*dc c(6) -1.29 ***(-3.00)

C c(7) 0.13 ***(29.04)

R2 0.74

F-Stat 26.84

Prob(F) 0.00

Durbin-Watson Stat. 1.04

Total Obs. 45 576

Wald Test H0: c(5)+c(6) = 0 -0.81 ***(-2.34)

Note: This table shows the results of estimating a VECM model for an unbalanced panel of 4780 publicly tradedcompanies from 55 countries and 9 sectors over the period 2004-12. Cross-section and time fixed effects are used inthe regressions as is the White cross-section covariance method. *, ** and *** indicate statistical significance at the10%, 5% and 1% levels, respectively.

y y y with Nt o t h t h t t t= + + + +− −α α α ξ ξ σ1 120~ ( , )

σ β β ξ β σφ ν

t o tt

t jt

21 1

2

12

2

1

= + +−=

−=

∑ ∑

∑ −β ξ1 12t

∑ −β σ22t j

σt2

problems in the international financial system - · pdf fileproblems in the international...

Documents